cturan/llama.cpp

spogulis no https://github.com/cturan/llama.cpp

Autors	SHA1 Ziņojums	Datums
Acly	638d330246 ggml : fix graph reallocation with multiple chunks (#16396)	3 mēneši atpakaļ
Acly	f2a789e334 ggml : split graph allocations according to backend max buffer size (#15815)	3 mēneši atpakaļ